You export a handful of columns from the MLS and call it comps, because that is what the standard report gives you. The full export carries hundreds of columns and both sets of remarks, and the private ones are where other agents wrote down what actually happened on those deals.
A title company educator is teaching a room of agents his three levels of AI, and at level two he stops on the MLS. His framing: "You're sitting on the data."
The workflow he shows is short. Run the search you want — his example is closed transactions in a neighborhood you are about to present in. Then: "We could download that data out of export option called full export." Then upload that spreadsheet and have the AI run the analysis. "we can have AI run an analysis on that area. We could run it for a neighborhood, a city, a county."
Why bother when market reports already exist: "we have access to all these these market reports, but sometimes they're not granular enough." They come back at city or ZIP level and you need one neighborhood. "And this is how you would do it with the most accurate real data is through your actual MLS login."
The part almost nobody thinks about is what is inside the file. He describes his own MLS export as one where "it's got like over 400 data columns," "which would normally take us hours to go through and analyze", and then the line that matters: "It's got all of the remarks, private, public, and it's going to analyze all of that very, very quickly with AI."
His stated use for it is the listing that is sitting: ask whether it is priced right, and ask "What are the unique nuances of this property compared to other comparable properties in that area".
“We could download that data out of export option called full export.”
Every agent complains that public market reports are too broad, then exports three columns from the one database that is not. The private remarks are the real find — concession notes, why a deal fell apart, what a listing agent told other agents and nobody else. A model reads all of it in seconds and tells you what your sitting listing has in common with the ones that never sold, which is a better answer for a seller than another opinion from you.
Two cautions. The column count is his description of his own MLS and it varies by system. And your MLS has rules about exporting data and about what may be done with confidential remarks — read them before the file leaves your computer, because analyzing it with AI is not a defense.